ASSESSING THE IMPACTS OF TRADE PROMOTION INTERVENTIONS: WHERE DO WE STAND?

Transcription

1 ASSESSING THE IMPACTS OF TRADE PROMOTION INTERVENTIONS: WHERE DO WE STAND? Christian Volpe Martincus Impact Evaluation of Trade-Related Policy Interventions: Paving the Way World Bank Washington, December 7, 2010

2 Outline Information Barriers and Trade Promotion The Effects of Trade Promotion: What Did We Know? Producing More Reliable Estimates The Effects of Trade Promotion: New Evidence What Should Be Taken into Account The Way Ahead

3 Information Barriers and Trade Promotion Relevant, accurate, and timely information is a key input to effective marketing decisions. This is particularly true for firms operating beyond national boundaries. Among other things, firms must know the export process at home, the different ways of shipping the merchandise and their associated costs, the potential markets abroad and their demand profile, the conditions for entering these markets, and channels available to raise awareness of their products and those through which these products can be marketed.

4 Information Barriers and Trade Promotion Firms pursuing cross-border economic opportunities must specifically engage in a costly process of identifying and assessing the reliability, trustworthiness, timeliness, and capabilities of potential trade partners. Information gaps limit the ability of firms to learn about international trading opportunities and find a suitable trade partner, and in this way negatively affect exports.

5 Information Barriers and Trade Promotion How important are information obstacles? Survey-based empirical studies indicate that lack of information is one of the most relevant export barriers, both in terms of frequency of occurrence and degree of severity. Most common export impediments are associated with identifying the initial contact and the marketing costs involved in doing business overseas. Econometric studies also provide evidence on the importance of information barriers (e.g., Rauch and Trindade (2002) on Chinese immigrant networks and Feenstra and Hanson (2004) on Hong Kong s intermediation of Chinese trade).

6 Information Barriers and Trade Promotion Information impediments have motivated public interventions, which are commonly called trade promotion actions. Economically, these actions might be and have been justified on the basis of market failures, primarily in the form of information externalities (Notice that, to be warranted, interventions would need to be welfare-improving, so cost need to be factored in ). Given that it is difficult to exclude third parties from information and that its use is non-rivalrous, there is a potential for free riding on the successful searches of firms for foreign buyers. Private returns from these exporting activities would be lower than the corresponding social returns, and investment in their development would then be sub-optimally low.

7 Information Barriers and Trade Promotion Trade promotion actions can help attenuate information problems. They can lower the fixed costs that firms incur when exporting for the first time and in entering specific new markets by reducing those costs associated with information gathering, e.g., carrying out overseas market studies on prices, product standards, and potential buyers. Trade promotion can then potentially facilitate the internationalization of companies and specifically their penetration into new country and/or product markets. But is this actually the case?

8 What Did We Know? Trade Promotion Organizations (TPOs) generally rely on client satisfaction surveys and calculations based on firm-level customs data to assess the effects of their actions. Surveys provide these organizations with qualitative indications on how they are doing. But the usefulness of this information is doubtful because evaluations based on non-objective data may be easily biased. Furthermore, these studies are generally carried out by professional evaluators who work on commission and risk losing future clients if they provoke strong criticism.

9 What Did We Know? Some surveys ask the managers of firms about the volume of incremental sales associated with the assistance received from the organization. These measures also have several weaknesses. Managers are likely to exaggerate the size of the payoff because that would increase chances that the program would continue. Surveys may suffer from being non-representative (the response rate tends to be markedly uneven and relatively low).

10 What Did We Know? TPOs with access to comprehensive firm-level customs data generally perform what can be called direct imputation. They directly take the sum of the values of exports or compute the change in this value for those firms that they have assisted, attributing the export outcomes of these firms--and the resulting expansion of national exports as well--as their contributions. These figures are likely to overestimate the impact of export promotion support as it is implicitly assumed that these foreign sales or their increment would not have taken place in the absence of this support. This is evidently a very questionable assumption.

11 What Did We Know? Until very recently, only a few studies in the empirical trade literature have used data at the firm-level to evaluate more rigorously the impact of public policies on firm export behavior. Álvarez and Crespi (2000) conclude that instruments managed by PROCHILE had a positive and direct effect on the number of destination markets to which firms export, and indirectly, after a period of four years, on product diversification. Álvarez (2004) show that, whereas PROCHILE s trade shows and trade missions do not significantly affect the probability that firms become permanent exporters, its exporter committees do.

12 What Did We Know? Bernard and Jensen (2004) find that average US states expenditures on export promotion per firm do not significantly influence the probability that they will export. Goerg et al. (2008) show that grants aimed at increasing investment in technology, training, and physical capital, when large enough, appear to be effective in increasing exports of Irish firms that are already exporting, but are not effective in encouraging new firms to enter international markets. While insightful, these studies concentrate just on the manufacturing sector and/or are based on small samples of firms. Further, they do not fully identify the specific channels through which export promotion may affect firm exports.

13 What Did We Know? In short, our understanding of the effects of trade promotion actions has been at least limited. There has been a clear need to provide export promotion organizations with a set of analytical tools to help them evaluate these actions and better allocate funds across them to maximize their effectiveness.

14 Producing More Reliable Estimates Assessing the impact of public programs is essentially a counterfactual analysis. Causal inference about the effect of these programs requires determining how participants would have performed if they had not participated. This other outcome is counterfactual and unobservable by definition, as is the difference between a firm s exports if it uses the services provided by the TPO relative to what its exports would be in the absence of these services. Thus, the counterfactual outcome must somehow be recovered from the data available.

15 Producing More Reliable Estimates The common practice is to use non-supported firms as a control group. However, there may be systematic differences between supported and non-supported firms that may explain export performance. These difference should be accounted for to avoid biased impact estimates. Alternative methods have been proposed in the literature to control for firms differing characteristics and to construct the correct sample counterpart for the missing information on the outcomes had the firms not received services when no randomized groups are available (e.g.. Blundell and Costa Dias, 2009; and Imbens and Wooldridge, 2009).

16 Producing More Reliable Estimates In order to assess the effects of trade promotion on firms export outcomes, we used: Difference-in-differences Matching combined with difference-in-differences (SP and MP) Endogenous switching binary response models (Aakvik et al., 2005) Semi-parametric method for estimating quantile treatment effects (combined with difference-in-differences) (Firpo, 2007) And, in robustness exercises, GMM dynamic panel estimators and double-robust estimation, among others.

17 Producing More Reliable Estimates The methods described above have been applied on firm-level export data from six Latin American countries: Peru, Costa Rica, Uruguay, Chile, Argentina, and Colombia. For each country, the dataset consists of two main databases. The first database has highly disaggregated export data at the firm level for four to eight years (depending on the country) over the period from the national customs agencies.

18 Producing More Reliable Estimates Data are reported annually at the firm-product-country level to reveal how much a given firm exported of a certain product to a certain destination market in a particular year. Each record includes a firm s identifier, the product code (8 to 10- digit HS), the country of destination, and the export value in US dollars. In all cases the sum of these firms exports virtually adds up to the total merchandise exports as reported by the national central banks or the countries national statistical offices. Hence, these datasets cover the whole population of exporters and not merely a sample of manufacturing firms.

19 Producing More Reliable Estimates The second database consists of lists of firms assisted in each year of the respective sample periods, kindly provided by the export promotion organizations in aforementioned countries. One organization, PROEXPORT, has additionally furnished a list of companies using each of its main services. Finally, additional data have been gathered on exporters, such as employment and location (e.g., Peru, Costa Rica, and Argentina), starting data (e.g., Peru), and sales (e.g., Chile).

20 Producing More Reliable Estimates To notice: The main aim of the assessment has been to estimate the direct effect of trade promotion support on firms export outcomes, and not to evaluate it from a social welfare point of view. Since unfortunately required data were not available, the way export assistance affects other dimensions of firms performance such as productivity, total sales, or profits, could not be examined. For the same reason, it was not possible to analyze the impact of trade promotion activities on the overall firm extensive margin (i.e., the number of exporters).

21 Producing More Reliable Estimates The report in which the results of these assessments are presented as well as the technical background papers in which they are based can be found at:

22 New Evidence Peru: Extensive or Intensive Margin? Trade promotion actions have been associated with a higher rate of growth of firms total exports, primarily through increased rates of expansion of number of destinations and number of products, which precisely implies dealing with more severe information problems.

23 New Evidence Costa Rica: What Kind of Trade Does Trade Promotion Promote? Assisted firms already exporting only differentiated goods--whose trade faces greater information problems--had higher rates of growth of exports and number of destinations. No other significant effects were detected.

24 New Evidence Uruguay: Does Trade Promotion Help Firms Enter New Markets? Trade promotion has had a significant impact on the probability of adding a new country (LAC but not OECD). Export assistance does not appear to have had any impact on the probability that a firm would add new products in general, but only for differentiated goods, where information barriers to trade are higher.

25 New Evidence Chile: What are the Distributional Effects of Trade Promotion? Benefits from trade promotion programs seem to have been greater for smaller, relatively inexperienced firms as measured by their past total exports. These companies lack the scale and resources to assume the fixed costs associated with the penetration of new foreign markets and are those that face the greatest challenges in overcoming informational barriers.

26 New Evidence Argentina: What are the Distributional Effects of Trade Promotion? The positive effects of trade promotion programs on total exports and number of destinations have been stronger for small and mediumsize firms--as defined in terms of number of employees--, which are precisely those for which information problems tend to pose greater obstacles to trade.

27 New Evidence Colombia: Which Export Promotion Programs are Most Effective? Bundled services combining counseling, trade agenda, and trade missions and fairs, which provide exporters with comprehensive support throughout the process of starting export businesses and building up buyer-seller relationships with foreign partners, are more effective than isolated assistance actions.

28 New Evidence Recapitulating The size of the effects of trade support programs on firms export performance appears to be related to the severity of the information problems involved in the specific trading operation and/or faced by the firms carrying out this operation. In particular: Trade assistance effects are predictably greater on the extensive margin of firms exports (i.e., when firms attempt to increase the number of destinations or goods exported) and, specifically, when they seek to enter an entirely new country or product market. Trade promotion actions are more likely to generate larger export gains to the degree to which products traded are more differentiated.

29 New Evidence Recapitulating Relatively small firms with limited previous involvement in international markets can be expected to benefit more from trade assistance. Bundled support services provided throughout the export process from initiation of commercial contacts to establishment of business relationships seem to be more effective in enhancing firms export perspectives than individual actions. Do previous findings imply that larger companies seeking to expand their exports of reference-priced goods in their current destination markets should not be supported? Not necessarily.

30 What Should Be Taken into Account Current impact evaluation practices of many TPOs are flawed with methodological problems. These evaluations are likely to largely misrepresent the real contribution these organizations make to the companies export growth and thereby to that of their countries. In fact, previous estimates indicate that, over the sample period, the direct imputation strategy would on average overestimate these contributions in 5.9 times for PROMPERU; 9.4 times for PROCOMER; 7.3 times for URUGUAY XXI; 14.3 times for PROCHILE; 5.2 times for EXPORTAR; and 3.7 times for PROEXPORT. Hence, the outputs of these evaluations are not adequate for guiding the strategies and activities of these organizations and specifically the allocation of their generally scarce resources.

31 What Should Be Taken into Account The methods used to generate previous results certainly allow for substantial improvements in the accuracy of impact estimates, but, as almost any method, are not exempted from limitations. Identification is based on the assumption that time-varying unobserved effects do not affect selection into support and exports. There might be non-observed time-variant firm-specific factors that lead to improved export performance that are overrepresented in the group of assisted firms. If this is the case, then these procedures would overestimate the true causal effects of this assistance on export performance. Unfortunately, this possibility cannot be excluded. However, given the selection process into trade support, the extension of our sample period, and the covariates included, we would expect that, if any, this is not a too serious problem.

32 What Should Be Taken into Account In these evaluation exercises, general equilibrium effects are assumed away. However, if there exist information spillovers linked to participation in specific export promotion actions, then the outcome differences between assisted and non-assisted firms corrected by observable heterogeneity across these groups would underestimate the true impact of these actions. Under perfect contemporary dissemination of information across firms, this impact would not be statistically different from zero and could not be identified. According to informal tests, this does not seem to seriously threaten the validity of the estimation results reported. Nonetheless, these phenomena deserve to be explored more thoroughly in future research (as it does to what extent the size of the information spillovers justifies trade promotion interventions ).

33 What Should Be Taken into Account In some countries, firms may potentially get assistance from different public and private entities. Regrettably, in these cases there is no unified register of firms benefiting from various support measures. Thus, it is not actually possible to explicitly account for the influence of interventions other than trade promotion, with the result that these actions become an unobserved factor. If this factor is time invariant over the sample period, its impact will be automatically controlled for by the estimation procedures, which identify the effects of interest based on the time variation. If firms participation status in other assistance programs is instead a timevarying variable, we are back to the first scenario described above.

34 The Way Ahead The adequacy of the design of export promotion activities and the robustness of the evaluations of their effectiveness depend on the ability to access detailed information on the firms. Access of TPOs to relevant data held by other public agencies, including national bureaus of statistics, as well as collaboration to generate new relevant data should be improved. Strengthened cooperation between public and private entities would be desirable. The availability of richer and consistent databases would make it possible to go beyond the analysis of the effects of trade promotion actions on exports to include their impact on other measures of firm performance such as productivity.

35 The Way Ahead Databases including lists of beneficiaries of different programs would allow for carrying out reliable evaluations that take into account their mutual existence, assessing complementarities and synergies among them, and determining how to improve their coordination. Insights into these potential interdependencies would be a valuable input for designing policy instruments and defining their sequencing. The IDB is currently working along these lines in the framework of a recently launched project aimed at analyzing the interplay between trade and innovation policies. Social experiments can generate further and, under certain conditions, more robust insights into the effects of trade assistance, and therefore appear as the natural next strategy to pursue. The IDB is also working with some LAC TPOs in these evaluations.

36